Naturalistic observation involves simultaneous recording of behavioral events by independent, trained observer/coders. It has been used to evaluate interventions with families of high risk or delinquent youths and to formulate theoretical models of social deviance. Although many observation systems were designed for similar purposes, they differ with respect to behavioral content, sampling procedures, and scoring. This creates problems in comparing results from different studies. The primary objective of this project is to translate behavior scores from one observation system into scores that correspond to those that would have been obtained had another system been used. In this way, investigators could report their findings in terms of each other's key dependent variables in addition to their own. The ability to do so should enhance the cumulative nature of research in the area of family intervention. Other objectives of this project are: (a) to conduct a psychometric appraisal of five different family observation systems, (b) to generate a library of videotapes that can be used for observer training and as a data source, and (c) to develop a general model for improving the comparability of data generated through different forms of assessment. A number of integrated research activities are necessary to fulfill these objectives. They include (a) production of videotapes depicting family interaction, (b) coding of tapes using different observation systems, (c) data reduction and file generation, (d) interobserver agreement analyses, (e) reliability analyses, (f) assessment of convergent and discriminant validity using multimethod-multitrait matrices, (g) canonical correlation analyses, (h) regression analyses, (i) cross validation, and (j) dissemination of findings.